inverting an axis shouldn't affect the autoscaling setting

[WeatherGod]

invert_xaxis() and friends shouldn't blindly set the axis limits. The autoscaling state should be left unchanged. There may be other places in the codebase where this change should be made.

Closes #1553

[pelson]

Looks good to me.

Is this easy to test? If so, would you mind adding a simple test for this?

Cheers,

[WeatherGod]

Yes, I was just about to push up a commit that tests this, but then I ran into the build failure on my machine.

[dmcdougall]

@WeatherGod "The" build failure? The ones we've been seeing recently on Travis? The travis builds are succeeding now...

[WeatherGod]

No, the one I asked about on the dev mailing list a couple hours ago.
@mdboom thinks it has to do with using deprecated numpy api. I am going to
try rebuilding later today with numpy 1.7.

[dmcdougall]

Whew ok. That's a relief.

[mdboom]

As for "other places in the code that may need this update": I just did a quick grep. hist2d and spy both set the limits unconditionally like this -- I wonder what the implications of adding those are.

[WeatherGod]

Ok, looks like I accidentally picked up a couple of the cherry-picked commits in my last rebase. Hopefully that isn't too much of an issue.

I will look into spy and hist2d.

[WeatherGod]

wrt spy() and hist2d(), I have to question why they even need to explicitly set the limits. Maybe at the time, the autoscaling logic wasn't very good, but it is much more mature now. Perhaps we should update these to use autoscale_view()?

[efiring]

On 2012/12/04 8:41 AM, Benjamin Root wrote:

wrt spy() and hist2d(), I have to question why they even need to
explicitly set the limits. Maybe at the time, the autoscaling logic
wasn't very good, but it is much more mature now. Perhaps we should
update these to use autoscale_view()?

Whatever you do, make sure spy is setting the same tight limits as at
present. I really don't see any point in changing it; it is a very
special-purpose function, and it isn't broken. To me, it is logical
that it should work the way it does, turning off autoscaling by
explicitly setting its limits.

[WeatherGod]

We can certainly still do the same setting of the limits. I just suggest that the autoscaling state should be left alone.

[dmcdougall]

@WeatherGod Is this good to go or are you going to add hist2d and spy to this PR?

[WeatherGod]

I think I will leave spy and hist2d out. We will have to ponder it some more.

[dmcdougall]

@WeatherGod Ok. Are you needing to add anything or can I merge this?

[WeatherGod]

no, go ahead and merge.